STRESS TESTING BANKS

Til Schuermann
Oliver Wyman
Wharton Financial Institutions Center
First Draft: September 2011
This Draft: April 2012
This Print: April 17, 2012
Prepared for the Committee on Capital Markets Regulation

Abstract

How much capital and liquidity does a bank need to support its risk taking activities?
During the recent (and still ongoing) financial crisis, answers to this question using standard
approaches, e.g. regulatory capital ratios, were no longer credible, and thus broad-based
supervisory stress testing became the new tool. Bank balance sheets are notoriously opaque
and are susceptible to asset substitution (easy swapping of high risk for low risk assets), so
stress tests, tailored to the situation at hand, can provide clarity by openly disclosing details of
the results and approaches taken, allowing trust to be regained. With that trust re-established,
the cost-benefit of stress testing disclosures may tip away from bank-specific towards more
aggregated information. This still provides the market with unique information (supervisors,
after all, have access to proprietary bank data) without dis-incentivizing market participants
from producing private information and trading on it with all the downstream benefits of
information-rich prices and market discipline.
Keywords: capital requirements, leverage, systemic risk.
JEL Codes: G21, G28, G20.

Oliver Wyman and Wharton Financial Institutions Center; til.schuermann@oliverwyman.com. I would like to
thank Mark Flannery, Itay Goldstein, Bengt Holmstrom, Bill Janeway, Umit Kaya, Ugur Koyluoglu, Andy
Kuritzkes, John Lester, Clinton Lively, Hashem Pesaran, Brian Peters, Barry Schachter, Hal Scott, and members
of the Committee for helpful discussions and suggestions, and Cary Lin for helpful research assistance. All
errors remain mine, of course.

1. Introduction
There are three kinds of capital and liquidity: 1) the capital/liquidity you have; 2) the
capital/liquidity you need (to support your business activities); and 3) the capital/liquidity the
regulators think that you need.1 Stress testing, regulatory capital/liquidity and bank-internal (socalled economic capital/liquidity) models all seek to do the same thing: to assess the amount of
capital and liquidity needed to support the business activities of the financial institution. Capital
adequacy addresses the right side of the balance sheet (net worth), and liquidity the left side
(share of assets that are liquid, however defined). If all goes well, both economic and
regulatory capital/liquidity are less than the required regulatory minimum, and their difference
(between economic and regulatory) is small, namely that regulatory models do not deviate
substantially from internal model results.
Prior to their failure or near-failure, financial institutions such as Bear Stearns,
Washington Mutual, Fannie Mae, Freddie Mac, Lehman and Wachovia were adequately
capitalized, at least according to regulatory capital rules disclosed in their public filings.2 This
set of institutions spans a broad range of regulatory capital regimes and regulators: the SEC and
Basel 2 capital rules (Bear Stearns, Lehman), the OCC and the Federal Reserve and Basel 1
(Wachovia), the OTS (WaMu), and OFHEO (Fannie and Freddie) the last actually based on a
narrow stress scenario. All firms had broad exposure to residential real estate assets, either in the
form of whole loans (mortgages) or securities (MBS) or both, and all had internal risk models
which may or may not have deviated materially from the regulatory models (we dont know as
this is/was firm proprietary information). Yet to the question of what is the capital you need vs.
the capital you have, in each case the answer came out wrong. To be sure, neither firm-internal
(economic) or regulatory capital and liquidity models can guarantee failure prevention; indeed,
that is not their purpose as every firm accepts some probability of failure, sized by its risk
appetite. But the cascading of defaults, and the resulting deep skepticism of stated capital
adequacy by the market, forced regulators to turn to other tools for assessing, in a credible way,
the capital adequacy of banks. That tool turned out to be stress testing.3
This paper lays out a framework for the stress testing of banks: why is it useful and why
has it become such a popular tool for the regulatory community in the course of the recent
financial crisis; how is stress testing done design and execution; and finally, with stress testing
1

This pithy summary I owe to Peter Nakada.

Kuritzkes and Scott (2009) make the case for a more market-oriented assessment of capital adequacy.
3
Flannery (2012) argues that stress tests should be evaluated on a fair value (rather than book capital) basis.
2

results in hand, how should one handle their disclosure, and should it be different in crisis vs.
normal times. The framework is equally applicable to capital and liquidity adequacy, but for
simplicity the bulk of the discussion will focus on capital.
A successful macro-prudential stress testing program, particularly in a crisis, has at least
two components: first, a credible assessment of the capital strength of the tested institutions to
size the capital hole that needs to be filled, and second a credible way of filling that hole. The
U.S. bank stress test in 2009, the Supervisory Capital Assessment Program or SCAP, may serve
as a useful example. The U.S. entered 2009 with enormous uncertainty about the health of its
banking system. Absent more concrete and credible understanding of the problems on bank
balance sheets, investors were reluctant to commit capital, especially given the looming threat of
possible government dilution. With a credible assessment of losses under a sufficiently stressful
macroeconomic scenario, the supervisors hoped to draw a line in the sand for the markets: fill
this hole, and you wont risk being diluted later because the scenario wasnt tough enough.
Moreover, if some institutions could not convince investors to fill the hole, a U.S. government
program, namely Treasurys Capital Assistance Program (CAP), stood ready to supply the
needed capital. Importantly, the U.S. Treasury was a sufficiently credible debt issuer that the
CAP promise was itself credible.4 All banks with assets greater than $100bn (YE 2008) were
included, accounting for two-thirds of total assets and about half of total loans in the U.S.
banking system. In the end, ten of the 19 SCAP banks were required to raise a total of $75bn in
capital within six months, and indeed raised $77bn of Tier 1 common equity in that period.5
None needed to draw on CAP funds.
The European experience in 2010 and 2011 stands in stark contrast to the 2009 SCAP.
Against the background of a looming sovereign debt crisis in the peripheral euro-zone countries,
the Committee of European Bank Supervisors (CEBS) conducted a stress test of 91 European
banks in 2010 covering about two-thirds of total European bank assets and at least half in any
given participating country. The stress test included imposing haircuts on the market value of
sovereign bonds held in the trading book; the bulk of the sovereign exposure, however, was (and
is) in the banking book. Of the 91 banks, only seven were required to raise a total of 3.5bn
(< $5bn at the time) in capital. The level of disclosure provided was rather less than in the
SCAP. For instance, loss rates by firm were made available only for two sub-categories: overall
4

Note that the act of a sovereign recapitalizing its banks involves that sovereign issuing debt and then investing
(downstreaming) it as equity in the bank(s).
5
http://www.federalreserve.gov/bankinforeg/scap.htm.

retail and overall corporate.6 By contrast, the SCAP results released loss rates by major asset
class such as first-lien mortgages, credit cards, commercial real estate, and so on. Markets
reacted benignly nonetheless until a few months later when Ireland requested financial
assistance from the EU and the IMF. Subsequent stress tests of just the Irish banks revealed a
total capital need of 24bn; all had previously passed the CEBS stress test. Moreover, to help
close the credibility gap, the extent and degree of disclosure was far greater than any of the stress
testing exercises to date.7 The markets reacted favorably, with both bank and Irish sovereign
credit spreads tightening. The stakes for the 2011 European stress test, now conducted by the
successor to the CEBS the European Banking Authority (EBA) had risen substantially.
The results of 2011 EBA stress test of 90 banks in 21 countries were at first blush
similarly mild as the previous years.8 Eight banks were required to raise a total of only 2.5bn.
However, the degree of disclosure was much more extensive, nearly reaching the high bar set by
the Central Bank of Ireland in March 2011, including information on exposure by asset class by
geography. Importantly, all bank level results are available to download in spreadsheet form to
enable market analysts to easily impose their own loss rate assumptions. In this way the
official results were no longer so final: analysts could (and did) easily impose their own
sovereign haircuts on all exposures and thus test the solvency of any of the 90 institutions
themselves. A summary of the major macro-prudential stress tests to date is provided in Table 3,
and a summary of their disclosures in Table 1.

http://www.eba.europa.eu/EU-wide-stress-testing/2010/2010-EU-wide-stress-test-results.aspx.
http://www.centralbank.ie/regulation/industry-sectors/creditinstitutions/Documents/The%20Financial%20Measures%20Programme%20Report.pdf.
8
http://www.eba.europa.eu/EU-wide-stress-testing/2011/2011-EU-wide-stress-test-results.aspx.
7

Base &
Stress
Scenario

Bank level
results

Asset/Product
level loss rates

SCAP
March 2009

Stress

--

--

CEBS
July 2010

Both

Retail, all
corporate only

--

--

CCAR
March 2011

--

--

--

--

--

Ireland
March 2011

Both

Sovereign only

EBA
July 2011

Both

Retail,
corporate, CRE

High

--

--

--

--

CCAR
March 2012

Exposure detail
Bank vs.
(asset class,
supervisory/3rd
maturity, geography) party estimates

Table 1: Summary of disclosures across stress test exercises.

The SCAP was the first of the macro-prudential stress tests of this crisis. But the changes at
the micro-prudential or bank-specific level were at least equally significant, and they are
summarized in Table 2. With the SCAP, stress testing at banks went from mostly single (or a
handful) factor shock to using a broad macro scenario with market-wide stresses; from product
or business unit stress testing focusing mostly on losses to firm-wide and comprehensive,
encompassing losses, revenues and costs; all tied to a post-stress capital ratio to ensure a going
concern.
Pre-SCAP

Post-SCAP

Mostly single shock

Broad macro scenario and market

stress

Product or business unit level

Comprehensive, firm-wide

Static

Dynamic and path dependent

Not usually tied to capital adequacy

Explicit post-stress common equity

threshold

Losses only

Losses, revenues and costs

Table 2: Features of stress testing, pre- and post-SCAP

The remainder of the paper proceeds as follows. Section 2 briefly reviews the scant
literature, and Section 3 provides a discussion of how to design the stress scenario. Section 4
describes modeling approaches for the three components needed to implement stress testing:
losses, net revenues (profitability), and balance sheet dynamics. Section 5 reviews the disclosure
regimes across the different stress tests to date in more detail, presents a discussion of disclosure
in normal times, and Section 6 provides some concluding remarks.

2. Stress testing in the literature

Stress testing has been part of the risk managers toolkit for a long time. It is perhaps the
most basic of risk-based questions to want to know the resilience of an exposure to deteriorating
conditions, be it a single position or loan or a whole portfolio. Typically the stresses take the
form of sensitivities (spreads double, prices drop, volatilities rise) or scenarios (black Monday
1987, Fall of 1998, post-Lehman bankruptcy, severe recession, stagflation). These types of
stresses lend themselves naturally to understanding financial risks, particularly in a data rich
environment such as found in a trading operation. Nonfinancial risks like operational,
reputational and other business risks are much harder to quantify and parameterize yet rely
heavily on scenario analysis (earthquakes and other natural disasters, computer hacking, legal
risks, and so on). While the original Basel I Accord of 1988 did not make formal mention of
stress testing, with the Market Risk Amendment of 1995 stress testing merited its own section
and thus became embedded in the regulatory codex. Indeed evidence of stress testing
capabilities is a requirement for regulatory approval of internal models.
Risk management as a technical discipline came into its own with the publication of the
RiskMetrics technical document in 1994, and stress testing (of both kinds, sensitivities and
scenarios) is mentioned throughout. The first edition of Jorions standard-setting VaR book
(1996) had a subsection devoted to the topic it was elevated to a chapter in subsequent editions
and surely there are earlier examples. Stress testing as a risk management discipline was found
largely in the relatively data rich environment of the trading room, with the closely related
treasury function conducting interest rate scenarios and shocks.9 The Committee on Global
Financial Systems (CGFS) of the BIS conducted a survey on stress testing in 2000, and it
reinforces this view.10 In their summary of the CGFS report, Fender, Gibson and Mosser (2001)
9

See Kupiec (1999) and Berkowitz (2000) for more extensive discussions of VaR-based stress testing.
See CGFS (2000) and the summary of it principal findings in Fender, Gibson and Mosser (2001).

10

point out that most of the scenarios manifest in terms of shocks to market rates, prices or
volatilities. Typical examples are equity market crashes such as October 1987, rates shocks such
as 1994, credit spread widening such as during the fall of 1998, and so on. Such stress scenarios
have the virtue of being unambiguously articulated and defined and are thus transparent and easy
to implement and communicate on assets that have themselves natural market prices or
analogs, as is mostly the case in the trading book. More typical banking assets, such as corporate
loans (especially to privately held firms) and consumer loans (e.g. auto loans), are less naturally
amenable to this approach.
Formal stress testing of the banking book, which is dominated by credit risk, is more recent,
in part because quantitative credit risk modeling is itself a newer discipline.11 Perhaps stimulated
by the success of RiskMetrics, the late 1990s saw a spurt of activity in the development of credit
portfolio models, the two prominent examples being CreditMetrics (1997) and CreditRisk+
(Wilde, 1997).12 Stress testing, however, did not feature in these papers. Yet as Koyluoglu and
Hickman (1998) show quite clearly, all of these credit portfolio models share a common
framework of mapping outcomes in the real economy, often represented by an abstract state
vector, to the credit loss distribution, and thus should lend themselves naturally to stress testing.
With that in mind, Bangia et al. (2002), following broadly the CreditMetrics framework, show
how to use credit migration matrices to conduct macroeconomic stress tests on credit portfolios.
Foglia (2008) provides a survey of the literature (at least through late 2008) of stress testing
credit risk, both for individual banks or portfolios as well as banking systems. More recently,
Rebonato (2010) with his suggestively titled book Coherent Stress Testing (we return to the
problem of coherence below), argues for a Bayesian approach to financial stress testing, i.e. one
which is able to formally include expert knowledge in the stress testing design, with an emphasis
on exploring causal relations using Bayesian networks.
With few exceptions, regulatory requirements on stress testing were thin prior to the crisis,
though considerable expectations about stress testing capabilities were voiced in supervisory
guidance in the U.S. Examples include the Joint Policy Statement on Interest Rate Risk (SR 9613), guidance on counterparty credit risk (SR 99-0313), as well as country risk management (SR
02-05). But banks had significant discretion with regard to specific design and implementation
11

To be sure, the credit rating agencies, having been in the business of rating corporate bonds for nearly a century,
likely employ stress testing in their bond rating methodology, but old documentation to this effect is hard to come
by.
12
For an excellent overview and comparison of these and related models, see Koyluoglu and Hickman (1998).
13
The most recent guidance on counterparty credit risk, SR 11-10, has greatly expanded on stress testing
expectations. All SR letter can be found at http://www.federalreserve.gov/bankinforeg/srletters/srletters.htm.

of their stress tests. Brian Peters, then head of risk in bank supervision at the New York Fed,
observed in March 2007 at an industry conference that no firm had a fully-developed program of
integrated stress testing that captured all major financial risks on a firm-wide basis.14 Market
risk stress tests were most advanced, while corporate or enterprise-wide stress testing, whereby
all businesses were subjected to a common set of stress scenarios, was at best in a developmental
phase.

3. Designing the Stress Scenario

One of the principal challenges faced by both the supervisors and the firms in designing
stress scenarios is coherence. The scenarios are inherently multi-factor: we seek to develop a
rich description of adverse states of the world in the form of several risk factors, be they financial
or real, taking on extreme yet coherent (or possible) values. It is not sufficient to specify only
high unemployment or only significant widening of credit spreads or only a sudden drop in
equity prices; when one risk factor moves significantly, the others dont stay fixed. The real
difficulty is in specifying a coherent joint outcome of all the relevant risk factors. For instance,
not all exchange rates can depreciate at once; some have to appreciate. A high inflation scenario
needs to account for likely monetary policy responses, such as an increase in the policy interest
rate. Every market shock scenario resulting in a flight from risky assets flight to quality
must have a (usually small) set of assets that can be considered safe havens. These are typically
government bonds from the safest sovereigns (e.g. U.S., Japan, Germany, Switzerland). To be
sure, as sovereign government budgets are increasingly strained, questioning the low-risk
assumption of those treasury instruments would certainly be a worthwhile stress scenario, but it
would need to define an alternative risk-free asset class to which capital can flee.
While the problem of coherence is generic to scenario design, it is especially acute when
considering stress scenarios for market risk, i.e. for portfolios of traded securities and
derivatives. These portfolios are typically marked to market as a matter of course and risk
managed in the context of a value-at-risk (VaR) system. Practically this means that the hundreds
of thousands (or more) positions in the trading book are mapped to tens of thousands of risk
factors, and those risk factors are tracked on a (usually) daily basis and form the data used to
estimate risk parameters like volatilities and correlations. Finding coherent outcomes in such a
high dimensional space, short of resorting to historical realizations, is daunting indeed.
14

Presentation delivered at Marcus Evans conference Implementing Stress Tests into the Risk Management
Process, Washington DC, March 1-2, 2007.

Compounding the problem is the challenge of finding a scenario where the real and the
financial factors are jointly coherent. The 2009 SCAP had a rather simple scenario specification.
The state space had but three dimensions GDP growth, unemployment, and house price index
(HPI) and the market risk scenario was based in historical experience: an instantaneous risk
factor impact reflecting changes from June 30 to December 31, 2008. This period represented a
massive flight to quality, the markets experienced the failure of at least one global financial
institution (Lehman), and risk premia at the time arguably placed a significant probability on the
kind of adverse real economic outcome painted by the tri-variate SCAP scenario. This solution
achieved a loose coherence of the real and financial stress. The price one pays for choosing a
historical scenario is the usual one: it does not test for something new. Figures 3 and 4 compare
some of these risk factors (real GDP, unemployment, equity and home prices indices) across the
three U.S. stress tests to date, both to each other as well as to actual realizations since 2008 Q4.
For the 2011 EBA test, the supervisors specified over 70 risk factors for the trading book,
eight macro-factors for each of 21 countries (macro-factors such as GDP growth, inflation,
unemployment, real estate price indices residential and commercial, short and long term
government rates, and stock prices), plus sovereign haircuts across seven maturity buckets. The
macroeconomic stress scenario was generated by economists at the ECB with reference to the
EU Commission baseline economic forecast.
All supervisory stress tests to date have imposed the same scenario on all banks. Naturally,
any scenario may be especially severe for some banks and much less so for others, depending on
the business mix and geographic footprint. This one-size-fits-all approach is analogous to the
problem of regulatory vs. internal economic capital models: the former by design is the same for
all banks, while the latter, being bespoke to a given bank, directly takes account of the particular
business mix of that bank. This problem of same vs. bespoke stress scenario becomes especially
acute when we move from crisis times, when there may be less debate about what a relevant
adverse scenario might look like, to normal times. The 2011and 2012 CCAR recognized this
problem and asked banks to submit results using their own scenarios (baseline and stress) in
addition to results under the common supervisory stress scenario. This was an important step
forward from the 2009 SCAP: by asking banks to develop their own stress scenario(s), which
was to reveal the particular sensitivities and vulnerabilities of their portfolio and business mix,
supervisors could learn from the banks about what they thought to be the high risk scenarios.
This is useful not just for micro-prudential supervision learning about the risk of a given bank
but also for macro-prudential supervision by allowing for the possibility of learning about
8

common risks across banks hitherto undiscovered or under-emphasized. With this dual
approach, supervisors could directly compare results across banks from the common scenario
without sacrificing risk-discovery.

4. Executing on the stress scenario: losses and revenues

With the macro-scenario in hand, how does one arrive at the corresponding micro-outcomes:
losses and revenues under adverse market and macroeconomic conditions? To date there is very
little discussion in the public domain on how to solve this problem, except perhaps for stress
testing the trading book. Indeed, one of the more important contributions of the supervisory
stress tests in the U.S. and Europe has been the accompanying methodology documents disclosed
by the supervisors which are, understandably, more heavily focused on the banking book.15
4.1. Modeling losses
For a firm active in many markets (product and geography), the first task is to map from the
few macro-factors into the many intermediate risk factors that drive losses for particular products
by geography. The EBA was forced to confront the geographic heterogeneity problem directly
by virtue of spanning 21 sovereign nations with rather different economies. U.S. supervisors,
stress testing an economic region just somewhat smaller than that of the EBA, left the task of
accounting for the not inconsiderable geographic heterogeneity to individual firms. Regional
differences are critical in modeling losses for real estate lending (residential and commercial) but
is hardly limited to those products. Since the U.S. experiences regional business cycles the
national business cycle obscures considerable variation across states nearly all lending has
some geographic component. For example, credit card losses are especially sensitive to
unemployment, and in July 2011, with the national rate at 9.1%, the state-level unemployment
rate ranged from 3.3% in North Dakota to 12.9% in Nevada. Similar dynamics are at work in
wholesale lending, particularly for SME (small and medium enterprise) lending whose
performance has a strong geographic component.
The problem of mapping from macro to more intermediate risk factors is not limited to
geography. An interesting example is auto lending and leasing where the collateral assets are
used cars. While auto sales invariable decline in a recession, and the decline in 2008-2009 was
unprecedented in the post-war period, used car sales typically suffer less. Yes, households buy
15

For SCAP, see http://www.federalreserve.gov/bankinforeg/bcreg20090424a1.pdf. For EBA, see

http://www.eba.europa.eu/EU-wide-stress-testing/2011/The-EBA-publishes-details-of-its-stress-test-scena.aspx.
For 2011 and 2012 CCAR, see http://www.federalreserve.gov/newsevents/press/bcreg/bcreg20110318a1.pdf and
http://www.federalreserve.gov/newsevents/press/bcreg/bcreg20120313a1.pdf respectively.

fewer cars in a recession, but if they do need to purchase a car, it is relatively more likely to be a
used car. So even if the default rate on auto loans increases significantly during a recession, the
corresponding loss given default (LGD) or loss severity need not. A useful indicator of the
health of the used car market, and thus the collateral of an auto lending portfolio, is the Manheim
index. Over the course of the most recent recession (Dec. 2007 June 2009), the index rose 4%
while total new auto and light truck sales declined by 37%.
The problem of loose coupling of loss severity to the business cycle is not limited to auto
loans. Acharya et al. (2007) show that for corporate credit, an important determinant of LGD is
whether the industry of the defaulted firm is in distress at the time of default. The authors make
a compelling asset specificity argument: if the airline industry is in distress, and a bank is stuck
with the collateral on defaulted aircraft loans or leases, it will be hard to sell those aircraft except
at very depressed prices. The healthcare sector may be relatively robust at that time, as indeed it
has been in the recent recession, but it is difficult to transform an airplane into a hospital.
The EBA disclosure on methodology is especially rich. In the March 2011 document, for
example, detailed guidance is provided on stressed probabilities of default (PDs) and stressed
LGDs. Note that such guidance presumes that a bank has implemented an internal credit rating
system for its commercial loan portfolio. For a Basel II bank this may not be unreasonable since
internal ratings, mapped to a common external scale such as those used by the rating agencies,
are a cornerstone of the Accord. With a credit rating (internal or external) in hand, computing
stressed default rates for the portfolio becomes a straightforward exercise, either by assigning
higher PDs to a given rating, or by imposing a downward migration on the current portfolio.16
Since the EBA stress test was based on risk weighted assets (RWA) computed using Basel II risk
weights which are ratings sensitive, banks were forced to make use of stress migration matrices
to compute not only increased defaults (the last column of the matrix) but also the entire future
ratings distribution to arrive at the correct RWA value. The U.S. stress tests were conducted
under Basel I risk weights which are not obligor ratings sensitive. The fuss about RWA
calculations matters since the denominator of capital ratios, used to determine whether or not a
bank needs to raise capital, is RWA. To be sure, this complicates any comparison of U.S. and
European stress test results.

16

Of the 90 participating banks, 59 were so called IRB (internal ratings based) banks, meaning their internal models
were validated to the supervisors satisfaction for at least one regulatory portfolio (e.g. corporate, commercial real
estate, etc.). Non-IRB banks were given very non-specific guidance (EBA 2011a, Section 5.5.1.1).

10

Implementation in the trading book is more straightforward and has a rich discussion in the
public domain; see inter alia Allen, Boudoukh, and Saunders (2004), Jorion (2007), or
Rebbonato (2010). In a nutshell, existing positions are simply repriced using the stress scenario
risk factors, subject to the proviso that the risk factor mapping problem, discussed in Section 3,
has been solved. The corresponding problem of stressing the counterparty credit risk that comes
with derivatives activities has received less attention.17 Counterparty credit risk arises when, in a
derivative transaction revalued to the stress scenario, the bank finds itself in the money (i.e.
enjoys a derivative receivable) yet cannot be sure that the counterparty to the transaction will be
solvent to make good on the payment. Thus the value is discounted, where the discount is a
function of the expected default likelihood of the counterparty under the stress scenario, which
presumably is higher than today. This adjustment is called a credit value adjustment (CVA), and
banks with significant derivative activities manage CVA as a matter of course. As Canabarro
(2010) and Hopper (2010) point out, the modeling challenge to stress testing counterparty credit
risk is considerable. Not only does the PD of the counterparty change in a stressful environment,
but so does the exposure. Thus any CVA stress test involves two distinct simulation exercises.
If the collateral posted by the counterparty is anything other than cash or cash equivalent, a
revaluation of that collateral under the same stress scenario needs to be added to the process.18
4.2. Modeling revenues
Implementing stress scenarios on the revenue side of the equation remains largely a black
box and seems far less well developed than stress testing for losses. Neither the 2009 SCAP nor
the otherwise richly documented 2011 EBA disclosures devoted much space nor revealed much
detail about the methods and approaches for computing revenues under stressful conditions.
Total income in banks can be roughly divided into interest and non-interest income. Interest
income is clearly a function of the yield curve and credit spreads posited under the stress
scenario, but what the net impact of rising or falling rates are on bank profitability remains
ambiguous, perhaps in part because of interest rate hedging strategies (English 2002,
Purnanandam 2007). The impact of stress scenarios on noninterest income, which includes
service charges, fiduciary, fees, and other income (e.g. from trading), is far harder to assess, and
there is precious little discussion of its determinants in the literature. This is concerning since
Stiroh (2004) shows that not only has the share of noninterest income been steadily rising in U.S.
17

For an excellent treatment, see Canabarro (2010) and Hopper (2010).

There is the added complication that major derivatives dealers actively manage CVA risk using a range of
strategies and instruments that themselves vary in price and availability depending on market conditions.

18

11

banks, from 25% in 1985

1
to 43%
% in 2001, bu
ut it is associiated with grreater volatillity and loweer
usted returns.. If we comp
pare the 200
09 SCAP, thee 2011 EBA
A and the 20112 CCAR strress
risk-adju
tests, the median ban
nk in the U.S
S. was able to
o cover abouut 58% of itss total projeccted losses w
with
ncluding resserve releasees, if any) in 2009 and 633% in 2012,119 comparedd with 66% inn the
profits (in
European
n case. As Figure
F
1 show
ws, there is considerable
c
e variability across banks, especiallyy in
the EBA test, where in some casees profits ev
ven under thee stress scenaario are projjected to outp
tpace
losses 4:1!

Figure 1: Projeccted coverag

ge of losses with profits in
n 2009 SCAP
P and 2011 EB
BA stress tessts.

19

PPNR caalculations in the

t 2012 CCAR
R were net of operational
o
riskk related lossess, OREO expennses, as well ass
mortgage repurchase
r
and
d put-back costs, meaning theese items were not reported seeparately (thouugh they totaledd
$115bn forr all 19 banks) (Board of Gov
vernors, 2012)..

12

4.3. Modeling the balance sheet

Recall that capital adequacy is defined in terms of a capital ratio, roughly capital over assets.
Of course both the numerator and denominator are nuanced. All supervisory stress tests have
insisted to varying degrees that the relevant form of capital be common equity. The 2010 CEBS
test allowed for some forms of hybrid capital typical of state participations, but the requirements
were tightened a year later. As discussed in Section 4.1, the denominator is typically riskweighted assets (RWA), where the risk weights are determined by the prevailing regulatory
capital regime, namely Basel I (in the U.S. cases of the SCAP and CCAR) and Basel II (in the
two Europe-wide and the Irish stress tests). The many subtleties of what this implies is beyond
the scope of this paper, but suffice it to say that a bank may be forced to raise capital under one
regime but not the other, and without considerable detail about the portfolio, there is no way to
know which regime will result in a more favorable treatment.
Regardless of the risk weight regime, determining post-stress capital adequacy requires
modeling both the income statement and the balance sheet, both flows and stocks, over the
course of the stress test horizon, which is typically two years.20 This is illustrated in Figure 2
below. The point of departure is the current balance sheet, at which point the bank meets the
required capital (and, if included, liquidity) ratios. The starting balance sheet generates the first
quarters income and loss, which in turn determines the quarter-end balance sheet. The modeler
is then faced with the problem of considering the nature and amount of new assets originated
and/or sold during the quarter, and any other capital depleting or conserving actions such as
acquisitions or spin-offs, dividend changes or share (re-)purchase or issuance programs,
including employee stock and stock option programs. The problem of balance sheet modeling
exists under a static (be it in raw form, as in the 2011 EBA, or in risk weighted form, as in the
2009 SCAP) or dynamic balance sheet assumption. The bank should not drop below the
required capital (and liquidity) ratios in any quarter. Moreover, at the end of the stress horizon,
the bank needs to estimate the amount of reserves needed to cover expected losses on loans and
leases for the following year. In this way the stress tests are really three years (or T+1 years for a
T-year stress test).

20

The horizon is 9 quarters for the CCAR as it is based on Q3, not Q4, balance sheets.

13

Two year dynamic forecast

Starting
balance
sheet

Q1 income
statement

Q1 end
balance
sheet

L
A

Q2 income
statement

L
P&L

Q8 end
balance
sheet

Q2 end
balance
sheet

P&L

L
A

Capital
and
liquidity
ratios

Capital
and
liquidity
ratios

Capital
and
liquidity
ratios

Capital
and
liquidity
ratios

Figure 2: Stress testing balance sheet and income statement dynamics.

5. Stress testing disclosure

Stress testing is here to stay, whether because it is just good risk management practice, or
because it is enshrined in legislation (through the Dodd-Frank Act). In the debate on the
disclosure regime, it is not clear that more is always better. We divide the discussion into crisis
and non-crisis or normal times, with the simple point that normal times may not require or even
desire the same degree of transparency as is clearly needed in times of crisis.
We have seen very large differences in disclosure across the different supervisory stress tests,
as summarized in Table 1. The SCAP in 2009 opened Pandoras box by disclosing projected
stress losses for each of the 19 participating banks, for eight different categories or asset classes,
as well as resources other than capital to absorb losses (mostly pre-provision net revenue and
reserve releases, if any). Until then, regulatory disclosures (e.g. Y-9C reports for U.S. bank
holding companies) reported only realized losses (the past), not projected losses (a possible
future). This allowed the market to easily check the severity of the stress test, not just in terms of
the scenario, but much more importantly in terms of the resulting outcomes at the bank level.
Given the crisis of confidence prevalent in the market at the time, this amount of transparency
was crucial. Two years later, the CCAR displayed a radically different disclosure regime: only
the macro-scenario was published, but no bank level results. The only indication of bank level
outcomes were subsequent dividend and other capital actions announced by some banks: banks
allowed to raise their dividends were interpreted to have passed the stress test. The market
digested this meager information event without a hiccup.

14

Dodd-Frank, however, requires the Fed to disclosure results of regular stress testing, and with
the 2012 CCAR, and the accompanying rules (final and proposed21), we got a glimpse of what
regular disclosure might look like. The 2012 CCAR disclosed nearly the same level of detail as
the 2009 SCAP, namely bank-level loss rates and dollar losses by major regulatory asset class
(following the categories of the FR Y-9C bank holding company reports): first and second lien
mortgages, commercial and industrial (C&I) lending, CRE, credit cards, other consumer, and
other loans. In addition, the Fed reported dollar PPNR, gains/losses on the AFS/HTM securities
portfolio, as well as trading and counterparty losses for those firms who were required to conduct
the trading book stress.22 And, as with the 2009 SCAP, the numbers reported were supervisory
estimates, not the bank-own estimates of losses (and PPNR) under the stress scenario.
By contrast, the 2011 Irish and 2011 Europe-wide EBA stress tests, both of which disclosed
after the CCAR, were considerable in their detail, including comparison of bank and third-party
estimates of losses in the Irish case (revealing the bias any bank is likely to have when estimating
its own potential losses), and data in electronic, downloadable form in the EBA case. Ireland
especially was suffering from an acute credibility problem, having emerged in July 2010 from
the CEBS stress test with flying colors only to require massive external aid four months later.
This divergent experience between Europe and the U.S. provides some hints on how to
design a disclosure regime during normal times. The discussion on the benefits and costs of
stress test disclosures in Goldstein and Sapra (2012) will help us. They argue persuasively that
in a world with frictions and strategic environments, the benefits (better market discipline) may
not outweight the costs: banks may make poor portfolio choices designed to maximize the
chance of pasing the test (window dressing) and thereby give up longer term value; traders may
place too much weight on the public information of stress test disclosure and be disincentivized
to produce private information about the banks; and finally, with information content of market
prices now damaged, market disclipline is harmed, and supervisors will find market prices less
useful for policy decisions (micro- as well as macro-prudential).
To be sure, some disclosure is still preferable to no disclosure, and Goldstein and Sapra
propose disclosing aggregated but not necessarily bank-specific results, with sufficient
information about category outcomes (loss rates by major asset class, for instance). Aggregation
has the advantage of being less wrong since idiosyncratic errors in estimating bank conditions
under hypothesized stress scenarios are averaged away. In this way supervisors can still provide
21
22

http://www.gpo.gov/fdsys/pkg/FR-2011-12-01/pdf/2011-30665.pdf
In 2012, these were the six institutions with the largest trading portfolios.

15

useful macro-prudential information which only they can provide loss rates by asset class, total
capital decline in the system (or significant fraction of the banking system) without drowing
out signals about individual banks from the market participants themselves.
During times of crisis, with enormous uncertainty about the health of the banking system, the
ability of supervisors to correctly assess the health of individual firms, and the resulting inability
of the market to be able to tell a good bank from a bad, then the benefit of detailed bank-specific
stress test disclosure may very well be significant. Indeed Goldstein and Sapra argue that stress
test disclosures, when more disaggregated, ought to be accompanied by detailed descriptions of
the exposures of the banks. This is precisely what was done in the Irish bank stress test of 2011,
an acute case of loss of confidence (and subsequent regaining), as well as the 2011 EBA stress
test. Because credibility of European supervisors was rather low by that point, only with very
detailed disclosure, bank by bank, of their exposures by asset class, by country, by maturity
bucket, could the market do its own math and arrive at its own conclusions.
Between March 2009 and March 2011, the 19 SCAP banks had raised about $300bn in
capital, the S&P500 had increased by 65%, the economy was no longer in recession, and
arguably the supervisory agencies had regained credibility. The non-event of the non-disclosure
of the 2011 CCAR suggests that the market seemed content to live in a state of symmetric
ignorance, to borrow a term from Dang, Gorton and Holmstrom (2010). Of course this might
change should the economy receive another adverse shock, but until it does, it is not clear that an
EBA-like disclosure regime is necessarily desirable nor stability enhancing. Europe, by contrast,
is not out of the weeds yet (as of this writing). Yet even the EBA is not limitless with its
disclosure of the 2011 stress test results. It is worth noting that funding liquidity was also
stressed at banks, but without disclosing the results. Because liquidity positions are highly
dynamic and thus subject to rapid change, snapshot disclosure, especially with delay (the as-of
date for the 2011 EBA stress test was YE 2010), is unlikely to be informative at the time of
disclosure.23
Recall the discussion in the introduction: regulatory capital models (risk weighting), internal
economic capital models and stress testing all have the same goal, namely to determine the
amount of capital needed to support the business (risk taking) of the bank. Both regulatory and
economic capital models (and especially the former) evolve very slowly and thus have difficulty
adapting to financial innovation and rapidly changing macro conditions. Indeed, some of the
23

Reuters, Sept. 2, 2011, EBA wont seek disclosure of bank liquidity. Available at
http://www.reuters.com/article/2011/09/02/idUSL5E7K23PI20110902.

16

innovation is motivated by those slowly evolving, one-size-fits-all regulatory capital rules.

Moreover, bank balance sheets are notoriously opaque and subject to easy-to-hide asset
substitution (higher risk for lower risk assets); Morgan 2002. Stress tests, especially macroprudential supervisory stress tests, are by construction adapted to the then current environment
and bank portfolios. Between balance sheet opacity, asset substitution and regulatory arbitrage,
it is easy to see the value of a pop quiz in the form of bespoke stress testing (Acharya et al.
2011).

6. Conclusion
The problem of sizing the amount of capital needed to support the risk taking of a bank is not
new; but the use of broad-based supervisory stress tests for an entire banking system is. The first
use was in 2009 in the U.S., and its success there has made it the supervisory and risk
management hammer to deal with all nails. A critical component of the exercise is the disclosure
of the results. The reason stress testing became an imperative was precisely because existing
approaches that were publicly disclosed, such as regulatory capital ratios, were no longer
informative and heavily (if not entirely) discounted by the market. To regain credibility,
supervisory authorities needed to disclose enough to allow the market to check the math.
But broad-based supervisory stress testing has not been universally successful, as the 2010
and 2011 European experience has shown. Nor is it clear how useful such broad supervisory
stress testing with concomitant disclosure will be as a matter of routine. Its value in the crisis
was undoubtedly its pop quiz nature. It was sprung on the banks at short notice, and thus very
difficult to manipulate through careful pre-positioning, and it was tailored to the situation at
hand, genuinely revealing new information to all participants and the public. As a result, trust
was regained. Once trust has been re-established, the cost-benefit of stress testing disclosures
may tip away from bank-specific towards more aggregated information. This still provides the
market with unique information (supervisors, after all, have access to proprietary bank data)
without dis-incentivizing market participants from producing private information and trading on
it with all the downstream benefits of information-rich prices and market discipline.

17

References
Acharya, Viral V., Sreedhar T. Bharath, and Anand Srinivasan, 2007. Does Industry-wide
Distress Affect Defaulted Firms? Evidence from Creditor recoveries. Journal of Financial
Economics 85, 787-821.
Acharya, Viral, Hamid Mehran, Til Schuermann, and Anjan Thakor, 2011. Robust Capital
Regulation. Federal Reserve Bank of New York Staff Report No. 490.
Allen, Linda, Jacob Boudoukh, and Anthony Saunders, 2004. Understanding Market, Credit and
Operational Risk: The Value at Risk Approach. Blackwell: New York, NY.
Bangia, Anil, Francis X. Diebold, Andr Kronimus, Christian Schagen and Til Schuermann,
2002. Ratings Migration and the Business Cycle, With Applications to Credit Portfolio
Stress Testing. Journal of Banking & Finance 26:2/3, 235-264.
Berkowitz, Jeremy, 2000, A Coherent Framework for Stress Testing. Journal of Risk 2, 1-11.
Board of Governors of the Federal Reserve System, 2009a. The Supervisory Capital
Assessment Program: Design and Implementation. April 24, 2009.
http://www.federalreserve.gov/bankinforeg/bcreg20090424a1.pdf.
_____, 2009b. The Supervisory Capital Assessment Program: Overview of Results. May 7,
2009. http://www.federalreserve.gov/bankinforeg/bcreg20090507a1.pdf.
_____, 2011. Comprehensive Capital Analysis and Review: Objectives and Overview. March
18, 2011. http://www.federalreserve.gov/newsevents/press/bcreg/bcreg20110318a1.pdf.
_____, 2012. Comprehensive Capital Analysis and Review 2012: Methodology and Results
for Stress Scenario Projections. March 13, 2012.
http://www.federalreserve.gov/newsevents/press/bcreg/bcreg20120313a1.pdf.
Basel Committee on Banking Supervision, 2009, Principles for Sound Stress Testing Practices
and Supervision. Available at http://www.bis.org/publ/bcbs155.htm.
Canabarro, Eduardo, 2010. Pricing and Hedging Counterparty Risk: Lessons Re-Learned? ch.
6 in Eduardo Canabarro (ed.), Counterparty Credit Risk, Risk Books, London, UK.
Committee on the Global Financial System, 2001, A Survey of Stress Tests and Current
Practice at Major Financial Institutions. Available at http://www.bis.org/publ/cgfs18.htm.
Dang, Tri V., Gary Gorton and Bengt Holmstrom, 2010. Financial Crises and the Optimality of
Debt for Liquidity Provision. Working paper, available at
http://mfi.uchicago.edu/publications/papers/ignorance-crisis-and-the-optimality-of-debt-forliquidity-provision.pdf.
English, William B., 2002. Interest Rate Risk and Bank Net Interest Margins. BIS Quarterly
Review, December, 67-82.
European Banking Authority, 2011a. 2011 EU-Wide Stress Test: Methodological Note. 18
March 2011. Available at http://www.eba.europa.eu/EU-wide-stress-testing/2011/The-EBApublishes-details-of-its-stress-test-scena.aspx.
European Banking Authority, 2011b. European Banking Authority 2011 EU-Wide Stress Test
Aggregate Report. 15 July 2011. Available at http://www.eba.europa.eu/EU-wide-stresstesting/2011/2011-EU-wide-stress-test-results.aspx.

18

Fender, Ingo, Michael S. Gibson and Patricia C. Mosser, 2001. An International Survey of
Stress Tests. Federal Reserve Bank of New York Current Issues in Economics and Finance
7:10.
Flannery, Mark J., 2012, Measuring Equity Capital for Stress-Testing Large Financial
Institutions. Working paper ..
Foglia, Antonella, 2008. Stress Testing Credit Risk: A Survey of Authorities Approaches.
Banca dItalia Occasional Paper No. 37.
Goldstein, Itay and Haresh Sapra, 2012, Should Banks Stress Test Results be Disclosed? An
Analysis of the Costs and Benefits. Working paper..
Gupton, Gregory M, Christopher Finger and Mickey Bhatia, 1997, CreditMetricsTM
Technical Document, this version: April 2. J.P. Morgan. Available at
http://www.defaultrisk.com/_pdf6j4/creditmetrics_techdoc.pdf.
Hirtle, Beverly, Til Schuermann and Kevin Stiroh, 2009. Macroprudential Supervision of
Financial Institutions: Lessons from the SCAP. Federal Reserve Bank of New York Staff
Report No. 409.
Hopper, Gregory, 2010. Stress Testing and Scenario Analysis: Some Second Generation
Approaches ch. 11 in Eduardo Canabarro (ed.), Counterparty Credit Risk, Risk Books,
London, UK.
Jorion, Philippe, 2007. Value at Risk: The New Benchmark for Managing Financial Risk, 3rd
Edition. McGraw Hill: New York, NY.
Koyluoglu, H.Ugur and Andrew Hickman, 1998. Credit risk: Reconcilable differences. Risk 11
(10), 5662.
Kupiec, Paul H. (1998). Stress Testing in a Value at Risk Framework, Journal of Derivatives
6(1): 724.
Kuritzkes, Andrew and Hal Scott, 2009. Markets are the best judge of bank capital. Financial
Times, 23 September.
Kuritzkes, Andrew, Til Schuermann and Scott M. Weiner, 2005, Deposit Insurance and Risk
Management of the U.S. Banking System: What is the Loss Distribution Faced by the
FDIC? Journal of Financial Services Research 27(3), pp. 217-243.
Morgan, Donald P., 2002, Rating Banks: Risk and Uncertainty in an Opaque Industry,
American Economic Review 92(4), 874-888.
Pesaran, M. Hashem, Bjrn-Jakob Treutler, Til Schuermann and Scott M. Weiner, 2006.
Macroeconomic Dynamics and Credit Risk: A Global Perspective. Journal of Money,
Credit and Banking 38:5, 1211-1261.
Purnanandam, Amiyatosh, 2007. Interest Rate Risk Management at Commercial Banks: An
Empirical Investigation. Journal of Monetary Economics 54, 1769-1808.
Rebonato, Riccardo, 2010. Coherent Stress Testing: A Bayesian Approach to the Analysis of
Financial Stress. John Wiley & Sons: New York.
Stiroh, Kevin, 2004, Diversification in Banking: Is Noninterest Income the Answer? Journal
of Money, Credit and Banking 36:5, 83-882.

19

Wilde, Tom, 1997, CreditRisk+ A Credit Risk Management Framework. Available at

http://www.csfb.com/institutional/research/assets/creditrisk.pdf.

20

Target capital
ratio*

SCAP
March 2009

CEBS
July 2010

4% T1C

# of
participating
banks
19

6% T1
6% T1

Participation criteria
(total coverage)

All bank holding companies with at

least $100 bn total assets

Balance sheet
assumptions

Total required
capital raise (for
# of banks)

Risk types included:

Market, Credit,
Liquidity (funding),
Operational

Constant
RWA

$75 bn (19)

M,** C

Constant total
assets

3.5 bn (7)

M, C

none

--

M, C

(~2/3 of total banking assets)

91 (20
countries)

Largest banks in country until at

least 50% of total assets are included
(~2/3 of total banking assets)

CCAR
March 2011

5% T1C

19

Original SCAP-19

Ireland
March 2011

6% T1C

Largest banks not in wind-down

mode

Allowed for
balance sheet
shrinkage

24bn (4)

M, C, L, O

EBA
July 2011

5% T1C

Largest banks in country until at

least 50% of total assets are included

Constant total
assets

2.5 bn (8)

M, C, L***, O

none

--****

M, C, O

10.5% T1C (in

base)
90 (21
countries)

(~2/3 of total banking assets)

CCAR
March 2012

5% T1C
4% T1; 8% Total;
3-4% leverage

19

SCAP-19
An additional 11 BHCs with assets
>$50bn

Table 3: Summary of macroprudential stress tests to date

*:

T1: Tier 1 capital ratio; T1C: Tier 1 Common (or Core) capital ratio

**: Only banks with at least $100 bn in trading assets were required to conduct the market risk stress test
***: Liquidity risk was not directly assessed, though funding stresses were taken into account, especially as related to sovereign stress impacting
funding costs for financial institutions.
****: 4 of the 19 did not pass in the sense of having not gaining non-objection to their submitted capital plans.

21

Real GDP growth

Stress-test scenarios vs. recent historical observations

Unemployment rate
Stress-test scenarios vs. recent historical observations

6%

14%

4%

12%

2%

10%

0%
8%

-2%
6%

-4%
4%

-6%

2%

-8%
-10%

0%

10

12

Quarters from start of period

10

12

Quarters from start of period

CCAR 1 from 2010 Q4

CCAR 1 from 2010 Q4

CCAR 2 from 2011 Q3

CCAR 2 from 2011 Q3

SCAP "more adverse scenario" from 2008 Q4

SCAP "more adverse scenario" from 2008 Q4

Historical from 2008 Q4

Historical from 2008 Q4

Source: Fed, The Supervisory Capital Assessment Program: Design and Implementation, 24 April 2009; Fed, Comprehensive Capital Analysis and Review: Objectives and Overview, 18 March,
2011; Fed, Comprehensive Capital Review document and Capital Plan review 22 November 2011; Datastream

Figure 3: U.S. real GDP and unemployment scenarios compared

22

Dow Jones total stock market index level

Stress-test scenarios vs. recent historical observations

House Price Index

Stress-test scenarios vs. recent historical observations

16,000

160

14,000

140

12,000

120

10,000

100

8,000

80

6,000

60

4,000

40

2,000

20
0

0
0

10

12

10

12

Quarters from start of period

Quarters from start of period

CCAR 1 from 2010 Q4

CCAR 1 from 2010 Q4

CCAR 2 from 2011 Q3

CCAR 2 from 2011 Q3

SCAP "more adverse scenario" from 2008 Q4

Historical from 2008

Historical from 2008 Q4

Source: Fed, The Supervisory Capital Assessment Program: Design and Implementation, 24 April 2009; Fed, Comprehensive Capital Analysis and Review: Objectives and Overview, 18 March,
2011; Fed, Comprehensive Capital Review document and Capital Plan review 22 November 2011; Datastream

Figure 4: U.S. equity and house price indices compared

23